Differential sensitivity to learning from positive and negative outcomes in cocaine users

Altered sensitivity to positive and negative outcomes may be linked to the maladaptive choices characteristic of substance use disorders. Few studies have determined the distinct roles that positive and negative outcomes play in stimulus-response learning in cocaine users. The purpose of the present study was to investigate sensitivity to learning from positive and negative outcomes on a probabilistic learning task in cocaine users employing human laboratory and crowdsourcing techniques.

Individuals who reported cocaine use were recruited for a laboratory study (Experiment 1) or an online study on Amazon.com's Mechanical Turk (mTurk) (Experiment 2). All participants completed a feedback-based probabilistic learning task in which images were classified into categories (A versus B). Positive and negative outcomes were provided in a probabilistic manner on separate trials. Proportion of optimal responses and response times were recorded.

Active cocaine users were less sensitive to learning from positive relative to negative outcomes. These effects were consistent across image type and session in the laboratory sample. Similarly, reduced sensitivity to learning from positive outcomes was observed in cocaine users on mTurk. Control participants did not show suboptimal performance following positive or negative outcomes.

This study extends the limited research on feedback-based learning in drug users by demonstrating reduced sensitivity to positive outcomes in cocaine users recruited in the human laboratory and online. Future studies on the clinical significance and mechanisms underlying this bias are needed to understand its relevance as a target for intervention development.